Ranking-Based Feature Selection Method for Dynamic Belief Clustering

摘要

In this paper, we investigate the problem of dynamic belief clustering. The developed approach tackles the problem of updating the partition by decreasing the attribute set in an uncertain context. We propose a based-ranking feature selection method that allows us to preserve only the most relevant attributes. We deal with uncertainty related to attribute values, which is represented and managed through the Transferable Belief Model (TBM) concepts. The reported results showed that, in general, there is a beneficial effect of using the developed selection method to cluster dynamic feature set in comparison with the other static methods performing a complete reclustering.